The pandemic accelerated the already ongoing trend of remote broadcasting, this led to more and more AES67 users experimenting with Wide Area Networks (WANs) despite AES67 being originally designed for Local Area Networks (LAN). The AES SC-02-12M working group on AES67 development saw here a need of describing what to expect on WANs and issued the AES-R20-2021 report in September 2021 on best practices of using AES67 outside of the typical LAN use case. We will present typical problems that arise when doing transmissions over long distance and leased networks and how to solve them while keeping interoperability with ST2210 and AES67. Be it over the internet, over a privately owned WAN, or in a data center, some constrained and limitations are to be taken into consideration: How to keep multiple sites PTP synchronous? How to prevent the issues of an unavoidable packet loss? Multicast or unicast? How to handle Quality Of Service? While most of those solutions are use case dependant, they often need a compromise. We will explain the origin of those problems and propose solutions that allow streaming of AES67 beyond the LAN.
File Type:
pdf
Categories:
Cloud and WAN
Presenters :
Nicolas Sturmel - Merging Technologies
Year :
2022
dlp_document_download :
AES67 beyond the LANNicolas Sturmel, Merging Technologies ▪High quality contribution ▪Live concerts with musicians apart ▪Monitoring ▪Voice over in multiple languages ▪Because the cloud may be cheaper ?Usecases ▪High quality contribution ▪Live concerts with musicians apart ▪Monitoring ▪Voice over in multiple languages ▪Because the cloud may be cheaper ?Usecases With the increasing demand of WAN communications in AES67/ST2110-30 networks, mostly due to COVID-19, the AES SC-02-12M, standard comité group on AES67 development started a project to issue recommendations: ▪What can I do with my AES67 devices ? ▪What can I add to my network to strengthen the connection ? ▪What should manufacturers add to the equipments ? The report was published September 25th 2021AES-R20-2021 A lot of contributions from: - Academics - Manufacturers - System integrators - Users AES-R20-202115 months, 60 calls, and a lot of discussion on what the best practices may be. A lot of contributions from: - Academics - Manufacturers - System integrators - Users AES-R20-202115 months, 60 calls, and a lot of discussion on what the best practices may be. Best Practices What is beyond the LAN ? / $ 1 % ( Q G S R L Q W & R I / $ 1 $ 9 L U W X D O ( Q G S R L Q W ' : $ 1 U H D G J Z J Z J Z 9 3 1 , Q W H U Q H W O R Q J G L V W D Q F H O L Q H / $ 1 $ What is beyond the LAN ? / $ 1 % ( Q G S R L Q W & R I / $ 1 $ 9 L U W X D O ( Q G S R L Q W ' : $ 1 U H D G J Z J Z J Z 9 3 1 , Q W H U Q H W O R Q J G L V W D Q F H O L Q H / $ 1 $ What is beyond the LAN ? / $ 1 % ( Q G S R L Q W & R I / $ 1 $ 9 L U W X D O ( Q G S R L Q W ' : $ 1 U H D G J Z J Z J Z 9 3 1 , Q W H U Q H W O R Q J G L V W D Q F H O L Q H / $ 1 $ What is beyond the LAN ? / $ 1 % ( Q G S R L Q W & R I / $ 1 $ 9 L U W X D O ( Q G S R L Q W ' : $ 1 U H D G J Z J Z J Z 9 3 1 , Q W H U Q H W O R Q J G L V W D Q F H O L Q H / $ 1 $ What is beyond the LAN ? / $ 1 % ( Q G S R L Q W & R I / $ 1 $ 9 L U W X D O ( Q G S R L Q W ' : $ 1 U H D G J Z J Z J Z 9 3 1 , Q W H U Q H W O R Q J G L V W D Q F H O L Q H / $ 1 $ ▪LAN ▪Mostly L2 ▪Fast ▪Small (10-100m) ▪Self administratedWhat is beyond the LANSelf Administrability ▪LAN ▪Mostly L2 ▪Fast ▪Small (10-100m) ▪Self administratedWhat is beyond the LAN▪WAN ▪L3 ▪Speeds varies ▪Long distances (10-10,000km) ▪Protocol limitationsSelf Administrability ▪LAN ▪Mostly L2 ▪Fast ▪Small (10-100m) ▪Self administratedWhat is beyond the LAN▪WAN ▪L3 ▪Speeds varies ▪Long distances (10-10,000km) ▪Protocol limitations▪Cloud ▪L3++ ▪Fast ▪Small ▪Very few control on infrastructure ▪Sharing traffic ▪Sharing platforms (VM)Self Administrability ▪Scale of the network can be 10,000 times bigger or even more. ▪New it equipments will a role in the network, and they may not be designed for high real time traffic ▪Packets can be delayed or sent out of order (jitter) ▪Packets can be lost ▪« I know when data enters my ethernet network but not when (and if) i gets out »Size matters Delay and jitter ▪100000 times 125us is 12,5 seconds ▪it does not mean you have 12,5 seconds (one trip) of latency. ▪While on the LAN, 90% of the delay is due to PDV ( jitter ), on the WAN it is much less (down to a few % on awesome networks) ▪However, the geographic delay will be noticeable very quickly and will require mitigation.Geographic delay vs Variable delay / $ 1 : $ 1 J H R J U D S K L F G H O D 3 ' 9 P L Q L P X P S D F N H W G H O D P D [ L P X P S D F N H W G H O D ▪Light travels at approximately 200 000 km/s in a fiber ▪That’s 5ms for 1000km ! ▪On top of that, account for network topologyAt least 5ms every 1000km Example, through the internet from the cloud ▪Synchronisation between nodes using PTPv2 is a key feature of AES67 and ST2110 ▪Node can create the media clock (sampling frequency) with high phase accuracy ▪Node know when a received sample has been sent on the network and can estimate latency ▪However, classic PTP algorithm are not designed to handle large jitter. And PTP follower will hardly lock to the follower.PTP and Synchronisation ▪Use the syntonized mode (frequency only), who cares about phase anyway ? ▪Relax the expected accuracy (do you really need sub-millisecond sync when you are kilometers away ?) ▪Use NTP or any other clock ▪Use special boundary clocks with jitter rejection ( ! Meinberg) ▪Use synchronisation out of the network, like GNSS-PTP leaders.Synchronisation: solutions PTP with GPStiming and sync PTP with GPStiming and sync PTP with GPStiming and sync PTP with GPStiming and sync Loosing packets ▪Loosing a packet (or having it delayed over the configured link offset) is rare on a LAN, not an a WAN: network congestion, link failure... are bound to happen. ▪Here, AES67-2018 does not provide any answer, but ST2110-10 does by using ST2022-7, a multi-path redundancy. ▪Other techniques such as FEC, SRT or RIST can also be used stream reliability stream reliability Some companies are specialised in providing edge decides that will take RAVENNA streams and make sure that arrive safely on the other end of the pipe Using a gatewayTu n n e l i n g A practical proof on concept (Dec 2020)Stream transport via a SRT Gatway. ▪Using gateways to embed the media streams (e.g.: VPN) ▪Using SIP with the classical NAT traversal technique ▪Using unicast vs multicast (for both PTP and media) ▪VXLAN, LAG, Encryption, QoS…The Cloud and network contrains ConclusionWidely used protocols like RTP and SDP allow AES67 and ST2110 to be easily transported on a large network. But achieving high performance (low latency, high quality) requires care. Often, the solution depends on the performances of the underlying infrastructures Thank younsturmel@merging.com
Downloads:
9